The Fortran Basics. Handout Two February 10, 2006

Transcription

1 The Fortran Basics Handout Two February 10, 2006 A Fortran program consists of a sequential list of Fortran statements and constructs. A statement can be seen a continuous line of code, like b=a*a*a or INTEGER :: a,b. A construct is a group of statements that together form a specific task and are defined by the Fortran language. You will learn some Fortran constructs later in these notes. When defining the syntax of a Fortran statement or construct, in these notes, the following will be used Anything inside curly brackets, { }, will be a Fortran keyword. Anything inside angular brackets, <>, is the choice of the programmer Anything inside square brackets, [ ], will be optional to the statement. 1 Program Structure When you are writing a Fortran program you must be careful to adhere to the required structure of a Fortran program. Each program unit must have this structure. A program unit can be the main program, a subroutine, function or a module. So far you have only been introduced to the main program unit, so for now do not worry about the others. The required structure is explained with the help of your power3.f90 code, 1. PROGRAM power3 2.!*** program to calculate 3rd power of IMPLICIT NONE INTEGER :: a, b a = 5 9. b = a*a*a PRINT*, a,b END PROGRAM power3 1. The program unit header statement must come first, for example PROGRAM <program name> Line 1 in the power3 code. 2. Next follows the specification region, in this region you put any external references, data type declarations, interface blocks and the IMPLICIT NONE statement. Don t worry about the items listed in the above sentence that you have not met yet they will be explained later when required. Lines 4 6 in the power3 code. 1

2 3. Next follows the execution statements that follow the logical order of whatever you are trying to do. Lines 8 11 in the power3 code. 4. Finally the END statement to tell the compiler that there are no more lines to compile. For example END PROGRAM <program name> Line 13 in the power3 code. 2 Intrinsic Data Types and IMPLICIT NONE You have already seen the keyword INTEGER used in the power3 code you wrote earlier. It was used to declare variables of integer type. Fortran has five intrinsic data types. The word intrinsic means that they are defined within the Fortran language standard. The five data types are; INTEGER: For exact whole numbers. REAL (and DOUBLE PRECISION): For real numbers, variables declared as DOUBLE PRECISION will be more accurate than REAL variables. COMPLEX: For complex numbers of the form x+i y. CHARACTER: For characters and strings of characters. LOGICAL: For variables that have two possible values,.true. or.false.. For this course ALL data types you use in your code must be declared before you use them. To insure this is done ALL of you codes must contain the IMPLICIT NONE statement at the beginning of each program you write and as you will see later in the course it must be defined in other sections of code where necessary. 2.1 Numeric declarations (REAL,INTEGER & COMPLEX) The syntax of numeric declarations are of the form; {type}[,{attribute-list}] [::] variable[=<value>][,variable [=<value>],...] The type is of course one of the data types INTEGER, REAL... The attribute-list is an optional set of keywords that can alter the nature of a variable being declared, their are many attributes that can be given and they will be dealt with as we need them. So for example the following are valid examples of numeric variable declarations. NOTE how the :: is identified as being optional, however if there is a attribute-list OR a variable is being assigned an initial value then the :: is mandatory. I recommend that you use the :: in all your declarations as it does no harm and improves readability of your code. REAL :: angle INTEGER :: i,j,start=1 INTEGER :: a,b,c DOUBLE PRECISION :: acc COMPLEX :: plane INTEGER, PARAMETER :: dozen=12 REAL, PARAMETER :: pi= NOTE: The last declaration, in the above list, is an example of a variable being defined with an attribute. The PARAMETER attribute dictates that a constant is being defined. This means you can never change 2

3 its value, later in your code, from the one it is given during its declaration. If you do try to assign a value to data type declared with the attribute PARAMETER the compiler will flag an error causing compilation to fail. This is useful for any of your own constants and also common constants for example π. 2.2 Numeric Operators All the standard numeric operators exist in Fortran. Operators act on operands. Some have one operand and are termed monadic and some have two and are termed dyadic. Here is a table of the common Fortran numeric operators, they are numbered in order of precedence [1] being highest and at the top. [1] ** : Exponentiation operator, dyadic [2] / : Division operator, dyadic [2] * : Multiplicative operator, dyadic [3] - : minus operator monadic for negation [4] + : additive operator, dyadic [4] - : minus operator, dyadic If you are building a complicated expression you can use parentheses to change the default order that sub-expressions are evaluated making sure that they are being evaluated in the order you require. Example Consider the following expression, 8.0/4.0*3.0=6.0 : There is some confusion here as * and / have the same order of precedence, so which will Fortran evaluate first? When this occurs Fortran will operate a left to right rule so the left most expression in the conflict will be evaluated first. So 8.0/4.0 is calculated and then the result is multiplied by 3.0. To have Fortran calculate first 4.0*3.0 and divide the result into 8.0 you could make use of parenthesis and write 8.0/(4.0*3.0)= This left to right rule however does not apply to exponentiation. For example a**b**c, in this case a right to left rule is used to dictate the order of evaluation. 2.3 Character declarations (CHARACTER) The syntax of character declarations are of the form; CHARACTER [(LEN=<length>)][,{attribute-list}] [::] variable[=<value>][,variable[=<value>],...] This is similar to the numerical declaration, however, there is now an option (LEN=<length >) specifier. This length specifier allows the declaration of a character variable of more than one character. Consider the following character declarations. CHARACTER :: char1,char2,letter="i" CHARACTER, PARAMETER :: a="a",b="b",c="c" CHARACTER (LEN=6) :: class1="mt3501", class2="mt3502",class3="mt3503" CHARACTER (LEN=*), PARAMETER :: news1="man United will win the league!" CHARACTER (LEN=*), PARAMETER :: news2="scotland will win the next world cup!" NOTE: The last declaration has an * as its length specifier. This is because PARAMETER (only) declarations can have their length assumed by the compiler, so you do not have to state it for this case. 3

4 2.4 Logical declarations (LOGICAL) The syntax of numeric declarations are of the form; LOGICAL[,{attribute-list}] [::] variable[=<value>][,variable [=<value>],...] Some examples of logical type declarations. LOGICAL :: switch, gate LOGICAL :: ans1=.true. ans2=.false. NOTE: The logical type can have only two values,.true. or.false Valid variable names Notice that power3.f90 uses the variable names a and b. This raises the question what are valid variable names? A Fortran 90 name can have up to 31 characters. This can be any combination of letters (upper case or lower case), the ten digits 0-9 and the underscore character ; the first character must be a letter. Unlike UNIX, which is CASE sensitive, Fortran treats upper and lower case letters as identical. However I recommend that you always use the convention that variables you define are always lower case while Fortran key words (such as PROGRAM and REAL ) are in upper case. This makes it easier for people, including you, to read your programs. 3 Programing Style Always comment your code so you can understand it later and so others can make sense of it if they have to read it (including those who have to mark your work). Remember a comment is any text on a line following a!. Remember that comments are ignored by the compiler. Any Fortran 90 keywords and intrinsic functions such as SIN, COS and TAN should be in UPPER CASE letters and user created entities such as variable names should be in lower case. Indentation, this is essential for program readability. An indentation of two spaces should be applied program unit bodies and any statements inside fortran constructs should be indented a further two spaces. This will become clearer as you see example codes in the notes. Always use IMPLICIT NONE You should always choose the names of your data types (REAL, INTEGER etc.) so that they reflect their purpose in the code. For example you could call a constant π in your code pi. 4 Decision Making :: The IF Statement and Construct Decision making is an essential and powerful component of programing. 4.1 The If Statement One of the most common and simplest form of decision commands in Fortran is the IF statement. It takes the general form, IF (<logical expression>) <statement> A logical expression always reduces to either.true. or.false.. It is rather useless but the most 4

5 basic form of a logical expression would be simply the two logical constants.true. and.false., for example IF (.TRUE.) PRINT*,"Hello From Me" This statement will always of course print Hello From Me because.true. can only ever reduce to itself.true Relational Operators, Logical Operators & Logical Expressions Relational Operators Relational operators are used to test the relationship between two numeric operands for that specific operator. The relational operators are, < or.lt. : less than > or.gt. : greater than == or.eq. : identically equal to /= or.ne. : not equal to <= or.le. : less than or equal to >= or.ge. : greater than or equal to as mentioned above any expression involving these operators will reduce down to either.true. or.false.. Logical Operators Logical operators act on logical expressions to create a larger logical expression. The Fortran logical operators are listed in the table below and are in order of precedence, given in square brackets, from the top to the bottom. [1].NOT. : Negate the following operand [2].AND. :.TRUE. if both operands are.true. [3].OR. :.TRUE. if at least one operand is.true. [4].EQV. :.TRUE. if both operands reduce to the same [4].NEQV. :.TRUE. if both operands are different The order of precedence is important when dealing with logical expressions, however you can change 5

6 the order in which sub-expressions are evaluated by making use of parenthesis. Consider, Exercise One (I): Work out the values of ans1, ans2 and ans3. Check them by writing a small code and printing them out to screen using the fortran PRINT command. Create a handout2 directory and in it create a exercise1 directory. Write your code in the exercise1 directory and call it logical test.f90. Remember to compile use the command f90 -o logical test logical test.f90. LOGICAL :: tst1=.true., tst2=.true., tst3=.false. LOGICAL :: ans1,ans2,ans3 ans1=tst1.or. tst2.and. tst3 ans2=(tst1.or. tst2).and. tst3 ans3=tst1.or. (tst2.and. tst3) PRINT*, ans1,ans2,ans3 NOTE : In the ans1 expression the tst2.and. tst3 sub-expression is evaluated first! Now try the following exercise that uses logical expressions that involve both relational operators and logical operators. Exercise One (II): Work out the values of test1, test2 and test3. Check them by writing a small code and printing them out to screen. Do this in the exercise1 directory but write the code in a new file called logical test2.f90. REAL INTEGER LOGICAL LOGICAL :: height=1.85, weight=95.0 :: age=55 :: drinker=.true., employed=.true., smoker=.true. :: test1,test2,test3 If (age>50) Print*,"Over fifty" test1 = employed.and. (age<45) test2 = smoker.and. drinker.and. (height<2.00) test3 = (.NOT. smoker.or. (age<=55).and. (weight>50)).and. (height<=1.84) PRINT*,test1,test2,test3 6

7 4.3 The If Construct The IF construct is more flexible than the single IF statement. The IF construct can be seen as having three forms Form One The IF construct in its most simple form is IF (<condition>) THEN < statement 1> < statement 2> < etc > This form simply performs a single test and if the test result is true then a series (or a single) Fortran statements that lie in the body of the construct are executed. Form Two IF (<condition>) THEN <statements if condition is true> ELSE <statements if condition is false> This second form of the construct allows the program to take action on a series of Fortran statements if the the condition is false as well as if it is true. This is done by the addition of the ELSE keyword. Form Three IF (<condition1>) THEN <statements if condition1 is true> ELSEIF (<condition2>) THEN <statements if condition2 is true> ELSEIF (<condition3>) THEN <statements if condition3 is true> ELSEIF (<condition4>) THEN <statements if condition4 is true> ELSE <statements all conditions are false> The final form of the IF construct allows the programmer to produce a mutually exclusive list of options. NOTE there can be any number of ELSEIF branches in the construct. Also NOTE that the 7

8 ELSE is optional. Exercise Two: Create an exercise2 directory and in it write a program to calculate a the roots of a quadratic equation y = ax 2 + bx + c. Make use of the IF construct to categorise the quadratic into two real roots, one repeated real root, complex roots or no roots at all (ie. it is not a valid quadratic). To categorise the quadratic you will have to calculate the value of the discriminent b**2-4*a*c. Output the root(s) to the screen. Also calculate where the the quadratic has a turning point and the nature of this turning point and print these out to the screen. This is your first, in this course, real attempt at a meaningful Fortran code so it may seem a little daunting but spend some time thinking about how you are going to approach the problem. It can help to write down your ideas on paper first! Remember the roots of a quadratic are given by, b + b 2 4ac & b b 2 4ac. 2a 2a NOTE : For this exercise do NOT use the data type COMPLEX. You can solve the quadratic using only the REAL data type for all three cases of roots. As this is your first real attempt at a piece of code much of the code is given below. You are advised to make use of this and fill in the bits of missing code indicated by ****. PROGRAM quad!## Program to investigate a quadratic equation (y=a*x*x+b*x+c) IMPLICIT NONE!## Force explicit declaration of all variables!## Declare required variables **** IF (check) THEN! check must be.true. if "a" does not equal zero **** IF (discrim > 0) THEN!** If real solutions exist then enter construct **** ELSEIF (discrim == 0) THEN **** ELSE **** ELSE PRINT*,"This is not a valid quadratic" IF (check) THEN!## Calculate the quadratics turning point **** IF (a.lt. 0) THEN PRINT*,"Turning point is a maximum" ELSE PRINT*,"Turning point is a minimum" 8

9 5 Looping :: The DO Construct Computers are very powerful when dealing with repetitive tasks. To allow the programmer to implement this Fortran has a construct called the DO loop. It has a few different forms; 5.1 The Indexed DO Construct One form of looping in Fortran is the Indexed DO construct which has the form. DO <loop variable>=<expr1>,<expr2>[,<expr3>] <body statements> The loop variable runs from expression one in increments of expression three until it reaches the value expression two and then the loop terminates. Note that the three expressions could all be simply integer numbers. If the optional stride of the loop (<expr3>) is omitted then a stride of one is assumed. For example you should now calculate the 3rd power of all integers up to 8. This can be achieved by using a DO loop. Exercise Three In an exercise3 directory type the following code and save it as power3 loop.f90 compile and run it. PROGRAM power3_loop!*** 3rd power of a for 1 <= a <= 8 IMPLICIT NONE INTEGER :: a, b DO a = 1, 8 b = a*a*a PRINT*,a,"to the power of three = ",b END PROGRAM power3_loop The line DO a = 1, 8 tells the computer to repeat the section of the program between this line and the line. The part a = 1, 8 means that this section of code is to be repeated for a = 1, a = 2, a = 3... a = 8. Run the program and make sure that it works. NOTE that the indentation used inside the DO loop makes it easy to see where it starts and ends. 9

10 Exercise Four In an exercise4 directory copy your power3 loop.f90 from exercise three changing its name to factorial.f90 in the copying process. Edit it to create a code that will calculate the factorial value of the first eight integers and print them out to screen. The copy command you should use is below. fortran1 exercise4> cp../exercise3/power3 loop.f90 factorial.f The Conditional DO Loop Construct It is useful to have a looping construct that allows termination of the loop only when a certain condition has been met. For this situation Fortran has the DO - looping construct. DO <statement body> IF (<logical expression>) EXIT <statement body> The logic of this style is that the <statement body> is iterated through many times until the <logical expression> is true. 5.3 The DO WHILE Construct Fortran has the DO WHILE looping construct that does a very similar job to the conditional DO loop construct above. You should NOT use this DO WHILE form as the above conditional DO loop construct is more general and it is clear where the loop exits. The DO WHILE loop always exits at the top of the loop construct as this is where the test takes place. It is presented in these notes as you may well see it in some older codes. DO WHILE <condition> <body statements> The logic of this style is that the <statement body> is iterated through many times until the <condition> is true. 10

11 5.4 Nesting Do Loops It is often useful to nest Do loops. DO i=1,10 DO j=1,50 <body statements> Exercise Five In an exercise5 directory write a code that sums up 10*( )+11*( )...+19*( ) use two nested DO loops to do the calculation that iterate over and 20 29! 6 Numeric Arrays : Part One Arrays can be declared in Fortran and are in fact one of the most powerful features of the Fortran 90 language. A one dimensional numeric array is simply a list of numbers and could be used, for example, to represent a vector. A two dimensional array could be used to represent a matrix. Arrays for a given numeric data type are declared using the attribute DIMENSION(<array size>) in the attribute list of a numeric declaration statement. To declare an INTEGER array of one dimension, called intlist to hold thirty values use, INTEGER, DIMENSION(30) :: intlist!** 1D array to create a two dimensional array of real numbers with four elements in each extent of the array use, REAL, DIMENSION(4,4) :: mat1!** 2D array For example the factorial program could save the answers into an array. PROGRAM factorial IMPLICIT NONE REAL, DIMENSION(20) :: fact INTEGER :: n, nmax = 20!** set the first value fact(1) = 1.0!** calculate factorials DO n = 2, nmax fact(n) = fact(n-1) * REAL(n) PRINT*,fact END PROGRAM factorial 11

12 6.1 Printing Arrays to the screen You have seen how to print integers, reals etc. to the screen using the PRINT statement. If mat1 was a two dimensional array of real numbers representing a matrix then the command, PRINT*,mat1 would just print out all the elements of the matrix in a messy list to the screen. To print out the matrix in a structured form use the print statements PRINT*,mat1(1,:)!** Row one PRINT*,mat1(2,:)!** Row two PRINT*,mat1(3,:)!** Row three PRINT*,mat1(4,:)!** Row four This works because in the code mat1(1,:), the brackets have two arguments in them the 1 st is the number 1 and means reference only row one (the first dimension of the array), the 2 nd argument is a colon (:). The colon means reference EVERY element in the column (the second dimension of the array). Exercise Six : Write a program to create a two dimensional array with four elements in each dimension to represent a matrix. Set each of the matrix elements to be equal to the sum of its row index plus twice its column index. Hint: use a nested DO loop. Print out the matrix to the screen 12